I think we need to start planning to add a "top-level" data structure.

[Rob Browning]

Right now, the top level of your data is just an AccountGroup,
identical in structure to all of it's sub-account groups.  However, I
think we need to plan (soon) to add a real top-level structure that
can act as the "bundle" for all of the relevant, non-account, data
associated with your finances.  For example, I think you should have
to reference that structure to get at commodities info, rather than
having them be global, and I think that that structure should probably
refer to the GUID entity table.  (In the guid case, 99% of the time,
all of the entity table pointers should point to the same table, but
in some special cases, like when you're trying to open two different
files to compare them for equality, that's the "wrong thing(TM)".)
There are some other bits that should go in there as well, like the
recurring events that apply to that specific set of accounts, this
probably gives you the general idea...

Perhaps this top-level structure should also be the one that knows
about the dataset's "back-end", and perhaps we should merge in all of
the "Session" semantics wrt locking, renaming, cloning, etc.

I think this is likely to be important RSN.

-- 
Rob Browning <rlb@cs.utexas.edu> PGP=E80E0D04F521A094 532B97F5D64E3930